Five years ago, computer science professor Ed Felten challenged the music industry by cracking the codes designed to stop music from being copied. Two years ago, he wrote software to show that peer-to-peer networks couldn't be regulated. Now he's broken into another type of supposedly secure data: election results.
On Wednesday, Felten, along with computer science graduate students Alex Halderman '03 and Ariel Feldman, released a paper detailing flaws in the Accuvote-TS, an e-vote machine manufactured by Diebold, Inc.
"If someone had physical access to one of the machines for less than a minute, they could install vote-stealing software or infect it with a virus that would spread to other computers," Halderman said.
Felton criticized Diebold for failing to ensure the security of its e-vote machines.
"We think that there were not many precautions taken in the design of this system against insertion of malicious software, which is the kind of attack we've primarily looked at," Felten said. "It is much too easy for a moderately-skilled programmer to carry out this attack and steal votes in an election."
Felten and his team presented their findings to the computer science department Wednesday. To illustrate their results, they allowed audience members to vote in a mock presidential election between George Washington and Benedict Arnold. Arnold, perhaps the most notorious traitor in American history, emerged victorious every time, regardless of how the votes were actually cast. There was no way to detect that the election had been stolen.
The team timed the release of its paper carefully. They wanted to make their results public as early as possible before the November elections but decided to wait until after Tuesday's primary in Maryland, which utilized e-vote machines.
"We couldn't in good conscience just sit on the information while the elections were coming up," Felten said. "We felt like we had to give citizens and the public officials a chance to cope with this risk."
The team obtained its findings using a privately owned machine contributed by an anonymous donor in May. The trio worked together for the entire summer.
"It was a good opportunity both to substantiate a lot of the claims that have been made before and, in addition, to really further strengthen the arguments that these existing systems are insecure," Feldman said.
Diebold issued a press release criticizing the study. The company argued that manipulation of their machines in the way the paper describes is unfeasible. It also said that it has already corrected the problems Felten and his team identified.
But Felten disagreed with Diebold's assertion.
"The things they say they've fixed are not the problems we reported," Felten said. "As far as we can tell, the changes they're pointing to would not solve the problems ... The problems we report can only be fixed by redesigning the actual hardware on the machine."
Despite their assertion that e-vote machines are flawed, the research team has not lost hope that a reliable and secure electronic voting system can be designed.
Felten, Halderman and Feldman said security could be improved if electronic voting machines continue to create paper records.
"If the machine prints out a receipt that the voter can place into a ballot box, then you can later randomly check the votes cast on paper to see if the machine has tampered with the results," Halderman said. "I think that's the only way to monitor the election."
Creating a secure voting system also requires independent review of the technology, said Dan Boneh, professor of computer science and electrical engineering at Stanford University. Boneh is part of ACCURATE, a National Science Foundation-sponsored research center which works to improve voting technology.
"If we're going to have faith in the voting process, you want to make sure that the voting technology being used does what it's supposed to do," Boneh said. "The best way to certify that is obviously by having an open peer review of how these systems work, and what their designs for security [are] and what attacks on them are possible."
Felten said he hopes his team's work will encourage greater public discussion about voting security.
"Public officials are the ones who buy these machines and decide how they are used, and of course they answer to the public," Felton said. "Attention by publicity is valuable and necessary to get things to change."






